speech_recognition/FunASR
1
0
Fork 0
mirror of https://github.com/modelscope/FunASR synced 2025-09-15 14:48:36 +08:00
Code Issues Actions Packages Projects Releases Wiki Activity

sensevoice

Browse Source
This commit is contained in:
游雁 2024-06-20 19:35:55 +08:00
parent f76598d4a1
commit 491d2eae27
17 changed files with 191 additions and 130 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

0
examples/industrial_data_pretraining/deepspeed/ds_stage1.json → examples/deepspeed_conf/ds_stage1.json
View File

0
examples/industrial_data_pretraining/deepspeed/ds_stage2.json → examples/deepspeed_conf/ds_stage2.json
View File

0
examples/industrial_data_pretraining/deepspeed/ds_stage3.json → examples/deepspeed_conf/ds_stage3.json
View File

0
examples/industrial_data_pretraining/deepspeed/ds_z0_config.json → examples/deepspeed_conf/ds_z0_config.json
View File

0
examples/industrial_data_pretraining/deepspeed/ds_z2_config.json → examples/deepspeed_conf/ds_z2_config.json
View File

0
examples/industrial_data_pretraining/deepspeed/ds_z2_offload_config.json → examples/deepspeed_conf/ds_z2_offload_config.json
View File

0
examples/industrial_data_pretraining/deepspeed/ds_z3_config.json → examples/deepspeed_conf/ds_z3_config.json
View File

0
examples/industrial_data_pretraining/deepspeed/ds_z3_offload_config.json → examples/deepspeed_conf/ds_z3_offload_config.json
View File

31
examples/industrial_data_pretraining/bicif_paraformer/finetune.sh
View File

@ -1,6 +1,8 @@
# Copyright FunASR (https://github.com/alibaba-damo-academy/FunASR). All Rights Reserved.
# MIT License (https://opensource.org/licenses/MIT)
workspace=`pwd`
# method1, finetune from model hub
# which gpu to train or finetune
@ -45,21 +47,38 @@ log_file="${output_dir}/log.txt"
mkdir -p ${output_dir}
echo "log_file: ${log_file}"
torchrun \
--nnodes 1 \
--nproc_per_node ${gpu_num} \
../../../funasr/bin/train.py \
deepspeed_config=${workspace}../../ds_stage1.json
DISTRIBUTED_ARGS="
--nnodes ${WORLD_SIZE:-1} \
--nproc_per_node $gpu_num \
--node_rank ${RANK:-0} \
--master_addr ${MASTER_ADDR:-127.0.0.1} \
--master_port ${MASTER_PORT: 26669}
"
echo $DISTRIBUTED_ARGS
torchrun $DISTRIBUTED_ARGS \
../../../funasr/bin/train_ds.py \
++model="${model_name_or_model_dir}" \
++train_data_set_list="${train_data}" \
++valid_data_set_list="${val_data}" \
++dataset_conf.batch_size=20000 \
++dataset="AudioDataset" \
++dataset_conf.index_ds="IndexDSJsonl" \
++dataset_conf.data_split_num=1 \
++dataset_conf.batch_sampler="BatchSampler" \
++dataset_conf.batch_size=6000 \
++dataset_conf.sort_size=1024 \
++dataset_conf.batch_type="token" \
++dataset_conf.num_workers=4 \
++train_conf.max_epoch=50 \
++train_conf.log_interval=1 \
++train_conf.resume=false \
++train_conf.resume=true \
++train_conf.validate_interval=2000 \
++train_conf.save_checkpoint_interval=2000 \
++train_conf.keep_nbest_models=20 \
++train_conf.use_deepspeed=false \
++train_conf.deepspeed_config=${deepspeed_config} \
++optim_conf.lr=0.0002 \
++output_dir="${output_dir}" &> ${log_file}

32
examples/industrial_data_pretraining/contextual_paraformer/finetune.sh
View File

@ -1,6 +1,8 @@
# Copyright FunASR (https://github.com/alibaba-damo-academy/FunASR). All Rights Reserved.
# MIT License (https://opensource.org/licenses/MIT)
workspace=`pwd`
# method1, finetune from model hub
# which gpu to train or finetune
@ -46,22 +48,38 @@ log_file="${output_dir}/log.txt"
mkdir -p ${output_dir}
echo "log_file: ${log_file}"
torchrun \
--nnodes 1 \
--nproc_per_node ${gpu_num} \
../../../funasr/bin/train.py \
deepspeed_config=${workspace}../../ds_stage1.json
DISTRIBUTED_ARGS="
--nnodes ${WORLD_SIZE:-1} \
--nproc_per_node $gpu_num \
--node_rank ${RANK:-0} \
--master_addr ${MASTER_ADDR:-127.0.0.1} \
--master_port ${MASTER_PORT: 26669}
"
echo $DISTRIBUTED_ARGS
torchrun $DISTRIBUTED_ARGS \
../../../funasr/bin/train_ds.py \
++model="${model_name_or_model_dir}" \
++model_revision="${model_revision}" \
++train_data_set_list="${train_data}" \
++valid_data_set_list="${val_data}" \
++dataset_conf.batch_size=20000 \
++dataset="AudioDatasetHotword" \
++dataset_conf.index_ds="IndexDSJsonl" \
++dataset_conf.data_split_num=1 \
++dataset_conf.batch_sampler="BatchSampler" \
++dataset_conf.batch_size=6000 \
++dataset_conf.sort_size=1024 \
++dataset_conf.batch_type="token" \
++dataset_conf.num_workers=4 \
++train_conf.max_epoch=50 \
++train_conf.log_interval=1 \
++train_conf.resume=false \
++train_conf.resume=true \
++train_conf.validate_interval=2000 \
++train_conf.save_checkpoint_interval=2000 \
++train_conf.keep_nbest_models=20 \
++train_conf.use_deepspeed=false \
++train_conf.deepspeed_config=${deepspeed_config} \
++optim_conf.lr=0.0002 \
++output_dir="${output_dir}" &> ${log_file}

21
examples/industrial_data_pretraining/llm_asr/demo_train_or_finetune.sh
View File

@ -30,10 +30,20 @@ init_param="${output_dir}/model.pt"
mkdir -p ${output_dir}
echo "log_file: ${log_file}"
torchrun \
--nnodes 1 \
--nproc_per_node ${gpu_num} \
../../../funasr/bin/train.py \
deepspeed_config=${workspace}../../ds_stage1.json
DISTRIBUTED_ARGS="
--nnodes ${WORLD_SIZE:-1} \
--nproc_per_node $gpu_num \
--node_rank ${RANK:-0} \
--master_addr ${MASTER_ADDR:-127.0.0.1} \
--master_port ${MASTER_PORT: 26669}
"
echo $DISTRIBUTED_ARGS
torchrun $DISTRIBUTED_ARGS \
../../../funasr/bin/train_ds.py \
--config-path "${workspace}/conf" \
--config-name "${config}" \
++train_data_set_list="${train_data}" \
@ -41,6 +51,9 @@ torchrun \
++dataset_conf.batch_size=4 \
++dataset_conf.num_workers=4 \
++train_conf.max_epoch=15 \
++train_conf.use_deepspeed=false \
++train_conf.deepspeed_config=${deepspeed_config} \
++optim_conf.lr=0.0001 \
++init_param="${init_param}" \
++output_dir="${output_dir}" &> ${log_file} &

37
examples/industrial_data_pretraining/llm_asr/demo_train_or_finetune2.sh
View File

@ -30,18 +30,39 @@ init_param="${output_dir}/model.pt"
mkdir -p ${output_dir}
echo "log_file: ${log_file}"
torchrun \
--nnodes 1 \
--nproc_per_node ${gpu_num} \
../../../funasr/bin/train.py \
deepspeed_config=${workspace}../../ds_stage1.json
DISTRIBUTED_ARGS="
--nnodes ${WORLD_SIZE:-1} \
--nproc_per_node $gpu_num \
--node_rank ${RANK:-0} \
--master_addr ${MASTER_ADDR:-127.0.0.1} \
--master_port ${MASTER_PORT: 26669}
"
echo $DISTRIBUTED_ARGS
torchrun $DISTRIBUTED_ARGS \
../../../funasr/bin/train_ds.py \
--config-path "${workspace}/conf" \
--config-name "${config}" \
++train_data_set_list="${train_data}" \
++valid_data_set_list="${val_data}" \
++dataset_conf.batch_size=1 \
++dataset_conf.num_workers=0 \
++train_conf.max_epoch=15 \
++train_conf.save_checkpoint_interval=1000 \
++dataset_conf.data_split_num=1 \
++dataset_conf.batch_sampler="BatchSampler" \
++dataset_conf.batch_size=6000 \
++dataset_conf.sort_size=1024 \
++dataset_conf.batch_type="token" \
++dataset_conf.num_workers=4 \
++train_conf.max_epoch=50 \
++train_conf.log_interval=1 \
++train_conf.resume=true \
++train_conf.validate_interval=2000 \
++train_conf.save_checkpoint_interval=2000 \
++train_conf.keep_nbest_models=20 \
++train_conf.avg_nbest_model=10 \
++train_conf.use_deepspeed=false \
++train_conf.deepspeed_config=${deepspeed_config} \
++optim_conf.lr=0.0001 \
++init_param="${init_param}" \
++output_dir="${output_dir}" &> ${log_file} &

30
examples/industrial_data_pretraining/paraformer/finetune.sh
View File

@ -1,6 +1,7 @@
# Copyright FunASR (https://github.com/alibaba-damo-academy/FunASR). All Rights Reserved.
# MIT License (https://opensource.org/licenses/MIT)
workspace=`pwd`
# which gpu to train or finetune
export CUDA_VISIBLE_DEVICES="0,1"
@ -40,27 +41,42 @@ scp2jsonl \
output_dir="./outputs"
log_file="${output_dir}/log.txt"
deepspeed_config=${workspace}../../ds_stage1.json
mkdir -p ${output_dir}
echo "log_file: ${log_file}"
torchrun \
--nnodes 1 \
--node_rank 0 \
--nproc_per_node ${gpu_num} \
../../../funasr/bin/train.py \
DISTRIBUTED_ARGS="
--nnodes ${WORLD_SIZE:-1} \
--nproc_per_node $gpu_num \
--node_rank ${RANK:-0} \
--master_addr ${MASTER_ADDR:-127.0.0.1} \
--master_port ${MASTER_PORT: 26669}
"
echo $DISTRIBUTED_ARGS
torchrun $DISTRIBUTED_ARGS \
../../../funasr/bin/train_ds.py \
++model="${model_name_or_model_dir}" \
++train_data_set_list="${train_data}" \
++valid_data_set_list="${val_data}" \
++dataset_conf.batch_size=20000 \
++dataset="AudioDataset" \
++dataset_conf.index_ds="IndexDSJsonl" \
++dataset_conf.data_split_num=1 \
++dataset_conf.batch_sampler="BatchSampler" \
++dataset_conf.batch_size=6000 \
++dataset_conf.sort_size=1024 \
++dataset_conf.batch_type="token" \
++dataset_conf.num_workers=4 \
++train_conf.max_epoch=50 \
++train_conf.log_interval=1 \
++train_conf.resume=false \
++train_conf.resume=true \
++train_conf.validate_interval=2000 \
++train_conf.save_checkpoint_interval=2000 \
++train_conf.keep_nbest_models=20 \
++train_conf.avg_nbest_model=10 \
++train_conf.use_deepspeed=false \
++train_conf.deepspeed_config=${deepspeed_config} \
++optim_conf.lr=0.0002 \
++output_dir="${output_dir}" &> ${log_file}

75
examples/industrial_data_pretraining/paraformer/train_from_local.sh
View File

@ -1,75 +0,0 @@
# Copyright FunASR (https://github.com/alibaba-damo-academy/FunASR). All Rights Reserved.
# MIT License (https://opensource.org/licenses/MIT)
# method2, finetune from local model
workspace=`pwd`
echo "current path: ${workspace}" # /xxxx/funasr/examples/industrial_data_pretraining/paraformer
# download model
local_path_root=${workspace}/modelscope_models
mkdir -p ${local_path_root}
local_path=${local_path_root}/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch
git clone https://www.modelscope.cn/damo/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch.git ${local_path}
# which gpu to train or finetune
export CUDA_VISIBLE_DEVICES="0,1"
gpu_num=$(echo $CUDA_VISIBLE_DEVICES | awk -F "," '{print NF}')
# data dir, which contains: train.json, val.json
data_dir="../../../data/list"
train_data="${data_dir}/train.jsonl"
val_data="${data_dir}/val.jsonl"
# generate train.jsonl and val.jsonl from wav.scp and text.txt
python -m funasr.datasets.audio_datasets.scp2jsonl \
++scp_file_list='["../../../data/list/train_wav.scp", "../../../data/list/train_text.txt"]' \
++data_type_list='["source", "target"]' \
++jsonl_file_out="${train_data}"
python -m funasr.datasets.audio_datasets.scp2jsonl \
++scp_file_list='["../../../data/list/val_wav.scp", "../../../data/list/val_text.txt"]' \
++data_type_list='["source", "target"]' \
++jsonl_file_out="${val_data}"
tokens="${local_path}/tokens.json"
cmvn_file="${local_path}/am.mvn"
# output dir
output_dir="./outputs"
log_file="${output_dir}/log.txt"
config_name="config.yaml"
init_param="${local_path}/model.pt"
mkdir -p ${output_dir}
echo "log_file: ${log_file}"
torchrun \
--nnodes 1 \
--nproc_per_node ${gpu_num} \
../../../funasr/bin/train.py \
--config-path "${local_path}" \
--config-name "${config_name}" \
++train_data_set_list="${train_data}" \
++valid_data_set_list="${val_data}" \
++dataset_conf.batch_size=20000 \
++dataset_conf.batch_type="token" \
++dataset_conf.num_workers=4 \
++train_conf.max_epoch=50 \
++train_conf.log_interval=10 \
++train_conf.resume=false \
++train_conf.validate_interval=15 \
++train_conf.save_checkpoint_interval=15 \
++train_conf.keep_nbest_models=50 \
++optim_conf.lr=0.0002 \
++init_param="${init_param}" \
++tokenizer_conf.token_list="${tokens}" \
++frontend_conf.cmvn_file="${cmvn_file}" \
++output_dir="${output_dir}" &> ${log_file}

30
examples/industrial_data_pretraining/paraformer_streaming/finetune.sh
View File

@ -1,6 +1,7 @@
# Copyright FunASR (https://github.com/alibaba-damo-academy/FunASR). All Rights Reserved.
# MIT License (https://opensource.org/licenses/MIT)
workspace=`pwd`
# which gpu to train or finetune
export CUDA_VISIBLE_DEVICES="0,1"
@ -41,25 +42,42 @@ scp2jsonl \
output_dir="./outputs"
log_file="${output_dir}/log.txt"
deepspeed_config=${workspace}../../ds_stage1.json
mkdir -p ${output_dir}
echo "log_file: ${log_file}"
torchrun \
--nnodes 1 \
--nproc_per_node ${gpu_num} \
../../../funasr/bin/train.py \
DISTRIBUTED_ARGS="
--nnodes ${WORLD_SIZE:-1} \
--nproc_per_node $gpu_num \
--node_rank ${RANK:-0} \
--master_addr ${MASTER_ADDR:-127.0.0.1} \
--master_port ${MASTER_PORT: 26669}
"
echo $DISTRIBUTED_ARGS
torchrun $DISTRIBUTED_ARGS \
../../../funasr/bin/train_ds.py \
++model="${model_name_or_model_dir}" \
++train_data_set_list="${train_data}" \
++valid_data_set_list="${val_data}" \
++dataset_conf.batch_size=20000 \
++dataset="AudioDataset" \
++dataset_conf.index_ds="IndexDSJsonl" \
++dataset_conf.data_split_num=1 \
++dataset_conf.batch_sampler="BatchSampler" \
++dataset_conf.batch_size=6000 \
++dataset_conf.sort_size=1024 \
++dataset_conf.batch_type="token" \
++dataset_conf.num_workers=4 \
++train_conf.max_epoch=50 \
++train_conf.log_interval=1 \
++train_conf.resume=false \
++train_conf.resume=true \
++train_conf.validate_interval=2000 \
++train_conf.save_checkpoint_interval=2000 \
++train_conf.keep_nbest_models=20 \
++train_conf.avg_nbest_model=10 \
++train_conf.use_deepspeed=false \
++train_conf.deepspeed_config=${deepspeed_config} \
++optim_conf.lr=0.0002 \
++output_dir="${output_dir}" &> ${log_file}

29
examples/industrial_data_pretraining/seaco_paraformer/finetune.sh
View File

@ -1,6 +1,7 @@
# Copyright FunASR (https://github.com/alibaba-damo-academy/FunASR). All Rights Reserved.
# MIT License (https://opensource.org/licenses/MIT)
workspace=`pwd`
# which gpu to train or finetune
export CUDA_VISIBLE_DEVICES="0,1"
@ -44,22 +45,38 @@ log_file="${output_dir}/log.txt"
mkdir -p ${output_dir}
echo "log_file: ${log_file}"
torchrun \
--nnodes 1 \
--nproc_per_node ${gpu_num} \
../../../funasr/bin/train.py \
DISTRIBUTED_ARGS="
--nnodes ${WORLD_SIZE:-1} \
--nproc_per_node $gpu_num \
--node_rank ${RANK:-0} \
--master_addr ${MASTER_ADDR:-127.0.0.1} \
--master_port ${MASTER_PORT: 26669}
"
echo $DISTRIBUTED_ARGS
torchrun $DISTRIBUTED_ARGS \
../../../funasr/bin/train_ds.py \
++model="${model_name_or_model_dir}" \
++train_data_set_list="${train_data}" \
++valid_data_set_list="${val_data}" \
++dataset_conf.batch_size=20000 \
++dataset="AudioDatasetHotword" \
++dataset_conf.index_ds="IndexDSJsonl" \
++dataset_conf.data_split_num=1 \
++dataset_conf.batch_sampler="BatchSampler" \
++dataset_conf.batch_size=6000 \
++dataset_conf.sort_size=1024 \
++dataset_conf.batch_type="token" \
++dataset_conf.num_workers=4 \
++train_conf.max_epoch=50 \
++train_conf.log_interval=1 \
++train_conf.resume=false \
++train_conf.resume=true \
++train_conf.validate_interval=2000 \
++train_conf.save_checkpoint_interval=2000 \
++train_conf.avg_keep_nbest_models_type='loss' \
++train_conf.keep_nbest_models=20 \
++train_conf.avg_nbest_model=10 \
++train_conf.use_deepspeed=false \
++train_conf.deepspeed_config=${deepspeed_config} \
++optim_conf.lr=0.0002 \
++output_dir="${output_dir}" &> ${log_file}

36
examples/industrial_data_pretraining/sense_voice/finetune.sh
View File

@ -1,6 +1,7 @@
# Copyright FunASR (https://github.com/alibaba-damo-academy/FunASR). All Rights Reserved.
# MIT License (https://opensource.org/licenses/MIT)
workspace=`pwd`
# which gpu to train or finetune
export CUDA_VISIBLE_DEVICES="0"
@ -44,26 +45,39 @@ log_file="${output_dir}/log.txt"
mkdir -p ${output_dir}
echo "log_file: ${log_file}"
#torchrun \
#--nnodes 1 \
#--node_rank 0 \
#--nproc_per_node ${gpu_num} \
python \
../../../funasr/bin/train.py \
deepspeed_config=${workspace}../../ds_stage1.json
DISTRIBUTED_ARGS="
--nnodes ${WORLD_SIZE:-1} \
--nproc_per_node $gpu_num \
--node_rank ${RANK:-0} \
--master_addr ${MASTER_ADDR:-127.0.0.1} \
--master_port ${MASTER_PORT: 26669}
"
echo $DISTRIBUTED_ARGS
torchrun $DISTRIBUTED_ARGS \
../../../funasr/bin/train_ds.py \
++model="${model_name_or_model_dir}" \
++train_data_set_list="${train_data}" \
++valid_data_set_list="${val_data}" \
++dataset_conf.batch_size=500 \
++dataset="SenseVoiceDataset" \
++dataset_conf.IndexDSJsonl="IndexDSJsonl" \
++dataset_conf.data_split_num=1 \
++dataset_conf.batch_sampler="BatchSampler" \
++dataset_conf.batch_size=6000 \
++dataset_conf.sort_size=1024 \
++dataset_conf.batch_type="token" \
++dataset_conf.num_workers=0 \
++dataset_conf.num_workers=4 \
++train_conf.max_epoch=50 \
++train_conf.log_interval=1 \
++train_conf.resume=false \
++train_conf.resume=true \
++train_conf.validate_interval=2000 \
++train_conf.save_checkpoint_interval=2000 \
++train_conf.keep_nbest_models=20 \
++train_conf.avg_nbest_model=10 \
++train_conf.use_deepspeed=false \
++train_conf.deepspeed_config=${deepspeed_config} \
++optim_conf.lr=0.0002 \
++debug=true \
++device="cpu" \
++output_dir="${output_dir}" #&> ${log_file}